They Did It : The Effects of Emotionalized Blame Attribution in Populist Communication

Transcription

1 644026CRXXXX / Communication ResearchHameleers et al. research-article2016 Article They Did It : The Effects of Emotionalized Blame Attribution in Populist Communication Communication Research 2017, Vol. 44(6) The Author(s) 2016 Reprints and permissions: sagepub.com/journalspermissions.nav DOI: / journals.sagepub.com/home/crx Michael Hameleers 1, Linda Bos 1, and Claes H. de Vreese 1 Abstract How can we explain the persuasiveness of populist messages, and who are most susceptible to their effects? These questions remain largely unanswered in extant research. This study argues that populist messages are characterized by assigning blame to elites in an emotionalized way. As previous research pointed at the guiding influence of blame attributions and emotions on political attitudes, these message characteristics may explain populism s persuasiveness. An experiment using a national sample (N = 721) was conducted to provide insights into the effects of and mechanisms underlying populist blame attribution with regard to the European and national levels of governance. The results show that emotionalized blame attributions influence both blame perceptions and populist attitudes. Identity attachment moderates these effects: Emotionalized blame attributions have the strongest effects for citizens with weaker identity attachments. These insights allow us to understand how populist messages affect which citizens. Keywords political communication, (media) populism, responsibility attribution, populist blame attribution, populist attitudes Already more than a decade ago, Mudde (2004) spoke of a populist zeitgeist in Europe. If anything, this zeitgeist has become even more pervasive in recent years. In Greece, for example, the left-wing anti-establishment party Syriza has made it into the government. 1 University of Amsterdam, The Netherlands Corresponding Author: Michael Hameleers, The Amsterdam School of Communication Research ASCoR, University of Amsterdam, Nieuwe Achtergracht 166, 1018 WV Amsterdam, The Netherlands. m.hameleers@uva.nl

2 Hameleers et al. 871 Populism also prevails outside of Europe. The Tea Party in the United States, for example, is considered to be a platform for conservative populist discontent (Barstow, 2010). In the midst of the dissemination of the populist zeitgeist, a growing body of literature points to the pivotal role of communication in delivering the populist message to the people (e.g., Krämer, 2014; Mudde, 2004). Mudde even argues that the populist zeitgeist is partially caused by the media s receptivity to the populist discourse. The core of populism revolves around the moral distinction between the good people and the culprit elites, who are unable or unwilling to represent the people s will. This idea can be expressed by different actors, such as politicians, the media, or citizens. The media can be populist by framing issues in terms of the opposition between law-abiding citizens and culprit politicians. If citizens perceive an antagonism between the good people and the corrupt elites, their interpretation of societal issues can also be regarded as populist. Populist messages are argued to be highly persuasive as they respond to the ordinary people s hopes and fears while formulating easy solutions to important societal problems (Hawkins, 2010; Mudde, 2004; Rooduijn, 2014). However, previous research failed to demonstrate what specific elements of populist media messages contribute to their persuasiveness. Vasilopoulou, Halikiopoulou, and Exadaktylos (2013) do point to one key distinguishing feature of the communication tactic of populists: They attribute more blame to the government and the European Union (EU) than mainstream politicians. By highlighting the purity of the people and by referring to the establishment as culprit, populism is inherently about attributing blame to others while absolving the people of responsibility. Populist messages attribute blame using a highly emotionalized style, emphasizing anger and fear toward threatening political elites (Fieschi & Heywood, 2004; Mudde, 2004; Ruzza & Fella, 2011). Reasoned from this perspective, the populist core idea distinguishes itself from mainstream politics by emotionally blaming elites (e.g., Europe or national governments) for causing the problems of the heartland (Stanley, 2008; Vasilopoulou et al., 2013). Previous research on attributions of responsibility a concept strongly related to populist blame attribution indicates that citizens political attitudes are affected by messages that emphasize who should be blamed for causing political problems (Hobolt & Tilley, 2014). Moreover, it is shown that the emotional style of blame attributions bolsters their persuasiveness (Brader, 2005). Despite the fact that prior research implicitly hinted at the persuasiveness of populist blame attributions, no empirical studies on its effects and underlying mechanisms have been conducted. Therefore, the central aim of this study is to provide insight into how emotionalized blame attribution the core element of populist messages influences citizens blame perceptions and populist attitudes. We conducted a 3 2 between-subjects online survey experiment (N = 721) to assess how citizens blame perceptions and populist attitudes toward the elites of the national government and the EU are affected by emotionalized attributions of blame. Doing so, this study responds to calls by scholars who have emphasized that the effects of media populism on the receiver side of the populist discourse should be studied more thoroughly (Jagers & Walgrave, 2007; Krämer, 2014; Mazzoleni, 2008). In

3 872 Communication Research 44(6) general, we found that emotionalized blame attributions guide both blame perceptions and populist attitudes. Identification with the nation or Europe moderated these effects. These findings indicate that emotionalized blame attribution provides a relevant framework for understanding the persuasiveness of populist communication, which allows us to explain what message characteristics affect the populist attitudes of which citizens. Populist Communication as Emotionalized Blame Attribution Citizens are assumed to credit the government for successes and blame it for failures. This process, which is crucial for democratic functioning, is called attribution of responsibility (Gomez & Wilson, 2008; Malhotra & Kuo, 2008; Tilley & Hobolt, 2011). In this article, we focus on causal attributions of responsibility that attribute negative qualities to the elites on the national and European levels of governance drawing on the emotions of anger and fear. We therefore conceptualize these causal attributions of responsibility as emotionalized blame attributions to elites. But how is this blame game related to populism? Although many scholars still disagree on how to precisely define populism, consensus exists that populism entails the construction of a blameless in-group opposed to a culprit out-group (Canovan, 1999; Mudde, 2004; Taggart, 2000). As the moral opposition between the pure people against the corrupt elites is the lowest common denominator central to all types of populism, we take this populist core as a point of departure throughout this paper (Rooduijn, 2014). This core consists of two main components: (a) the perception that politicians should, but are failing to, represent the will of the ordinary people and (b) the perceived antagonism between the good people and the evil elites (Jagers & Walgrave, 2007). Populism can be connected to responsibility attribution. Populism always portrays the people as a homogeneous in-group belonging to the imagined community of the heartland (Taggart, 2000). This heartland is in a severe state of crisis because the elites (e.g., the EU and the government) failed to represent the will of the people (Canovan, 1999). Hence, the in-group is threatened by the corrupt elite, the homogeneous outgroup constructed in opposition to the people (Laclau, 1977; Mudde, 2004). As the people are silenced by corrupt politicians who are blind to see the real problems of the nation, the people cannot help their heartland being in a state of crisis: The corrupt elites are held responsible for causing the crisis of the heartland (Taggart, 2000). Attributing responsibility for negative outcomes blame to the corrupt elites for causing the people s problems is thus inherently populist (Jagers & Walgrave, 2007). In general, populist communication emphasizes the opposition of the people to representative democracy (e.g., Canovan, 1999). In Western Europe, this entails attributing blame to national as well as European levels of governance. For this reason, we focus on the EU and the national government as the culprit out-groups (Jagers & Walgrave, 2007; Taggart, 2000). An example of such blame-shifting rhetoric was articulated by the United Kingdom Independence Party (UKIP) leader Nigel Farage in

4 Hameleers et al. 873 an online British newspaper: Unless we leave the EU, a huge surge in population growth is unavoidable (Farage, 2015). In this message, the EU is blamed for causing a salient problem of the in-group of the British nation. The Role of the Media in Framing Populist Blame Attributions The media play a crucial role in disseminating populist messages to the people (Krämer, 2014; Mazzoleni, 2008; Muis & Scholte, 2013). When referring to the relationship between the media and populism, a distinction between populism by the media and populism intended for the media can be made (Bos & Brants, 2014). The latter form of relationship concerns the media s receptivity to populist ideas, whereas the former identified relationship taps into the concept of media populism (e.g., Krämer, 2014; Mazzoleni, 2008). Media populism can be defined as the adaptation of elements of populist ideas (e.g., construction of pure people versus corrupt elites) and style (e.g., emotionalized) by the media themselves (Krämer, 2014). In line with the concept of media populism, the media may actively use emotional blame attribution as a framework for the coverage of political issues. Using the populist core idea, the media emphasize the distinction between the culprit elites and the blameless people. Moreover, they communicate this message in an emotional style that reflects hostility toward these elites, for example, by emphasizing the outrage of the common people toward the corrupt government or by stressing the people s fear of the threatening EU. Emotionalized blame attribution is thus not only confined to populist actors, but it can also be part of the content of populist messages. The following quotes of messages posted on the Tea Party s website illustrate such content: Liz Cheney blasted Barack Obama s disastrous foreign policy and Barack Obama is a dangerous man and he is punch-drunk on power. Bash Obama on Facebook? Be afraid, very afraid ( Obama Has Proceeded Over Last 7 Years, 2015). In these messages, the U.S. president is not only personally blamed for the government s failing policy, the words blasted, disastrous, and afraid indicate that blame is attributed using an emotional style. The media can create effects of populism through framing, which entails the reconstruction of social reality in meaningful patterns of interpretation (Scheufele, 1999). In Entman s (1993) frequently cited definition of framing, media frames are conceptualized as consisting of four frame-elements: a problem definition, a causal interpretation, a moral evaluation, and/or a treatment recommendation. As populist messages emphasize who is causally responsible for the defined problem while morally evaluating the people s will as good and the elites influence as evil, the framing of populist messages connects most saliently to the first three frame-elements. The framing approach has previously been applied to research on both populism and attributions of responsibility. Assigning responsibility, which taps into the frameelement of causal interpretation, plays a central role in media framing as it affects citizens interpretation of political issues (Iyengar, 1991; Kühne, Weber, & Sommer, 2015). Reasoned from framing research on populism, the central populist frame has

5 874 Communication Research 44(6) previously been defined as us versus them (e.g., Caiani & Della Porta, 2011). This populist frame taps into the causal interpretation as well as the moral evaluation of issues as the in-group is perceived of as morally good and absolved of responsibility while they are evil and causally responsible. In line with this, Jagers and Walgrave (2007) argue that the populist master frame concerns the distinction between the blameless people and the corrupt elites. Synthesizing insights into the framing of populism and responsibility, we argue that the central populist frame should be perceived as a blame the corrupt elites who are opposed to the people frame. But what are the potential effects of these populist blame frames? As a consequence of exposure to messages that emphasize the divide between the people and the elites, citizens may interpret society in binary us against them oppositions themselves as well (Krämer, 2014). These populist interpretations are frequently expressed on social media, for example: The prime minister needs to go and we need out of EU now. By activating such interpretations among citizens, populist messages contribute to media based othering. This process can further be explained by psychological theories on stereotyping. Populist blame frames make negative stereotypes of the EU and the national government chronically accessible to the people (e.g., Dixon, 2008). Populist messages that accuse corrupt politicians for not providing enough jobs for ordinary hard-working citizens may, for example, strengthen citizens perception that they are not responsible for their own situation. This negative stereotype of the elites as culprit and the people as blameless victims becomes accessible when citizens think about politics. As political issues are difficult to comprehend and citizens frequently lack factual information on who is responsible for political issues at different levels of governance (e.g., Arceneaux, 2006; Cutler, 2004), citizens are expected to use the stereotypes provided in populist messages when forming an opinion on the political elites in the EU and the national government. As attribution of responsibility provides a powerful psychological cue for the formation of favorable attitudes toward the in-group and hostile attitudes toward the out-group, citizens blame perceptions and populist attitudes are expected to be guided by blame frames (Hobolt & Tilley, 2014; Krämer, 2014). By marking the moral distinction between us and them, populist blame frames are thus expected to affect citizens blame perceptions and populist attitudes. Against this backdrop, we hypothesize the following: Hypothesis 1 (H1): Blame attributions in populist communication guide citizens blame perceptions and populist attitudes. Social Identity: The Perceptual Screen Moderating the Effects of Blame Attribution Our central argument is that blame attribution is at the core of populist political communication. Hypothesis 1 suggests that citizens ceteris paribus follow suit when either the national government or the EU is blamed. However, populist blame attribution may not be effective for all citizens. Krämer (2014) expects that citizens who do

6 Hameleers et al. 875 not share the worldview articulated by populist media will respond to populist communication with reactance. Moreover, Lenz (2009) showed that individuals accept the ideas of parties they identify themselves with and reject the ideas of parties they oppose. In studies on attributions of responsibility, it is demonstrated that government partisans are less likely to accept blame attribution to the government than supporters of the opposition (Tilley & Hobolt, 2011). Likewise, citizens feelings of closeness to national or European identity may restrain them from forming negative attitudes toward the national government or the EU. But what is the underlying process that makes citizens resistant to persuasion by the populist blame game? In social psychology, the process underlying populist blame attribution, in-group versus out-group construction, is called social categorization (Aronson, Wilson, & Akert, 2007). This process results in out-group homogeneity and in-group serving bias. In-group bias is defined as the experience of positive feelings for the in-group, in this case the Dutch nation-state or the EU, and negative feelings for the out-group constructed as significantly different from the EU or the nation. Out-group homogeneity means that people belonging to the out-group are perceived as more similar to each other than they actually are (Aronson et al., 2007). We call these in-group and outgroup biases the perceptual screen. Social identity theory explains how the perceptual screen mechanism influences attitudes toward the in-group and out-group (Tajfel, 1978). This theory argues that the experience of belonging to the in-group forms a crucial part of people s self-concept. As people want to maintain a positive self-concept, they are biased in their judgments and attribute negative qualities (e.g., blame) to the out-group that is perceived as significantly different (Gordon & Arian, 2001; Tajfel, 1978). For a consistent positive self-concept, the in-group is absolved of blame (McLaren, 2007; Sniderman, Peri, de Figueiredo, & Piazza, 2000). In line with this mechanism, people are expected to only accept blame attributed on the national level when they do not feel close to their nation. When people do feel connected to their nation, accepting frames that attribute blame on the national level is inconsistent with their positive self-concept. Likewise, we expect that people will only blame the EU when they are not attached to European identity. These effects are explained by in-group serving bias, functioning as a frame of reference or a perceptual screen when citizens assign responsibility (Tilley & Hobolt, 2011). Attachment to identity may thus be key in understanding the conditionality of populist communication s effects. In this study s context, people who identify strongly with Dutch culture and citizenry should not be persuaded by populist messages that blame their government. If people feel strongly connected to European identity, the EU is part of their in-group and therefore absolved of causal responsibility for negative outcomes. In other words, Hypothesis 2 (H2): Identity attachment functions as a perceptual screen when citizens construct blame perceptions and populist attitudes in response to populist political communication.

7 876 Communication Research 44(6) Specifically, Hypothesis 2a (H2a): Citizens who feel attached to national identity are less likely to accept frames attributing blame to the national government than citizens who do not feel attached to national identity. Hypothesis 2b (H2b): Citizens who feel attached to European identity are less likely to accept frames attributing blame to the EU than citizens who do not feel attached to European identity. The Role of Anger and Fear in the Populist Blame-Game Previous research indicated that emotionally charged information affects citizens opinions in a different way than information that is presented without references to emotions (Gadarian, 2010). Therefore, it is important to study how the effects of populist blame attribution depend on the use of different negative emotions central in populism. Populists attribute blame using an emotional communication style (Fieschi & Heywood, 2004; Ruzza & Fella, 2011). By emphasizing that the enemies (e.g., the corrupt elites of the EU and the national government) are creeping upon the heartland (Finlay, 2007), the emotional style of populist messages instills a sense of threat on the people. This threat is communicated by using the negative emotions of anger and fear, for example, by emphasizing how ordinary citizens fear the EU s impact on their insecure future or by stressing outrage toward the government s failing policy. In research on responsibility attributions, expressions of anger and fear are found to affect blame perceptions in different ways (e.g., Nabi, 2003). Populist communication draws on anger to emphasize that the culprit elites are blocking the goals of the people (Grant & Brown, 1995; Ruzza & Fella, 2011). Fear is used to highlight uncertainty about the threatening future of the heartland, which is in a state of crisis because the corrupt elites failed to represent the people (Mols & Jetten, 2014). The negative emotions of anger and fear thus play different roles in the populist blame game. As these negative emotions are central in populist constructions of the elites, we will investigate how anger and fear as emotional styles affect citizens responses to populist blame attribution messages. Research into the role of discrete emotions in framing often draws on appraisal theory. This theory describes how emotions are experienced as a consequence of a person s subjective interpretation of a situation (Lerner & Keltner, 2001). This interpretation is referred to as an appraisal tendency or an appraisal pattern. According to appraisal theory, specific emotions elicit specific appraisal patterns, which in turn influence how information is processed (Nabi, 2003). Crucially, anger and fear as discrete emotional styles have different effects on how citizens process information on causal attributions of responsibility for negative outcomes, or blame (Han, Lerner, & Keltner, 2007). Specifically, anger s appraisal pattern stimulates heuristic processing, whereas fear s appraisal pattern stimulates systematic processing (Kim & Cameron, 2011).

8 Hameleers et al. 877 Fear s appraisal tendencies of uncertainty and uncontrollability construct the perception of a threat that needs to be dealt with. To resolve feelings of uncertainty and to avert the threat, new information in the environment is processed systematically (e.g., Major, 2011). As fear stimulates systematic processing of new information, fear is expected to result in a tendency to accept populist blame attributions (e.g., Brader, 2005). This expectation is in line with the findings of Gadarian (2010), who found that fear-inducing cues increased the likelihood of a threatening message s acceptance. In contrast to fear, anger s appraisal pattern creates the perception of certainty and controllability. As feelings of certainty and control do not motivate people to search for new information, anger stimulates heuristic processing. This results in citizens dependency on existing attitudes (Lerner & Keltner, 2001; Leventhal & Scherer, 1987). Reasoned from appraisal theory, we hypothesize the following: Hypothesis 3a (H3a): The emotionalized style of fear will lead to a stronger tendency to accept populist blame attribution than anger. In addition, we expect that the effects of anger and fear will be contingent upon the experienced level of identity attachment. As we already predicted that fear has a stronger impact than anger, and weaker attached citizens are most susceptible to persuasion, we further hypothesize the following: Hypothesis 3b (H3b): The effects of the emotional style of the message are moderated by identity attachment, so that specifically at lower levels of identity attachment, fear will lead to stronger blame perceptions and populist attitudes than anger. At higher levels of identity attachment, the emotional style of the message should not affect blame perceptions and populist attitudes. Finally, reasoned from the perceptual screen mechanism, the acceptance of blame frames should differ at different levels of identity attachment. Moreover, appraisal theory predicts that fear leads to a stronger tendency to accept attributions of responsibility than anger. Synthesizing these predictions, we raise the following research question (RQ1): Research Question 1 (RQ1): Is the acceptance of the fear blame frame highest among lower levels of social identity? Method Design To test how emotionalized attributions of blame affect blame perceptions and populist attitudes, we conducted a 3 (Causal responsibility attribution: the EU vs. the national government vs. no-responsibility attribution) 2 (Emotionalized style: anger vs. fear) between-subjects factorial design with control group. Respondents were randomly assigned to one of the six experimental conditions or the control group (see Table 1, for all conditions).

9 878 Communication Research 44(6) Table Between-Subjects Factorial Design With Control Group. Emotional style blame attribution Anger Fear The EU The Dutch government No-blame attribution Control group Group 1: The EU is attributed blame; anger is highlighted Group 2: The EU is attributed blame; fear is highlighted Group 3: The Dutch Group 4: The Dutch government is attributed government is attributed blame; anger is highlighted blame; fear is highlighted Group 5: No blame is attributed; Group 6: No blame is attributed; anger is highlighted fear is highlighted No-blame attribution and no emotionalized style Sample This study is based on a diverse sample of Dutch citizens recruited by Research Now. A filter question was used to assess whether participants were able to adequately respond to a question about a short text. 1 The final sample consisted of 721 participants who complied with the instruction. 2 The mean age of the participants was years (SD = 16.62), and 52.9% were female. In all, 46.6% of the sample was lower educated, 42.3% was higher educated, and 13.9% had a moderate level of education. Informed by a pilot study, we expected that 100 participants per condition would result in a desired power of The actual observed power of the analyses to detect differences between experimental conditions was 0.79 (α =.05). Procedure The experiment was conducted online. Participants first completed the procedure of informed consent. Upon agreement, participants completed the filter question. Only if they were not screened out, they completed a pretest asking for demographics, the items for moderating variables and control variables. Next, participants were randomly allocated to one of the experimental conditions and read an online newspaper article. After reading this article, participants completed a posttest questionnaire measuring the dependent variables. At the end of the questionnaire, manipulation check items for causal responsibility attributions and emotionalized style were assessed. On average, participants completed the survey in minutes, of which seconds were spent on reading the newspaper article. Upon completion, participants received a financial incentive from the research agency. Independent Variables and Stimulus Materials The stimuli consisted of a negatively valenced online newspaper article on the worsening Labour market situation. This issue was chosen as it has both a national and an EU political dimension (Kriesi et al., 2008). As this issue is salient in Dutch public and political debates, it provides an externally valid case for which politicians and citizens

10 Hameleers et al. 879 evaluate causal responsibility all the time. The article was framed in seven different ways depending on the level of the independent variables emotionalized style and blame attributions. All versions of the article are included in Appendix A. Emotionalized style was manipulated in terms of appraisal patterns and emotion words that indicate anger and fear (e.g., Kühne, 2014; Nabi, 2003). In the fear conditions, pessimism, uncontrollability, and responsibility attributions directed at institutions related to the out-group were emphasized. Protection against rather than punishment of responsible actors was emphasized. The emotion words tension, fear, afraid, and pessimistic were used as indicators of fear (see also Nabi, 2003). In the anger conditions, controllability, certainty, and causal responsibility attributions to concrete actors were emphasized. Approach behavior was emphasized by highlighting the need to punish the specific actors responsible for the worsening Labour market situation. The emotion words angry, outraged, punish, and frustrated were used to indicate the discrete emotion of anger. We thus used four emotionindication words in each emotionalized style condition. To enhance external validity, the stimuli were based on four existing newspaper articles retrieved from online versions of three Dutch newspapers (de Volkskrant, de Telegraaf, NRC.next). The constructed articles ranged between 136 and 166 words in length and had the same number of paragraphs. The layout, style, language, sources, and content of the stimuli matched existing online newspaper articles. In all versions of the newspaper article, the storyline was identical and focused on how the Labour market situation has worsened over the last few years. The source, statistics, and references to the in-group of Dutch citizens were held constant between all experimental conditions. In each article, only one source was quoted: the governmental organization that deals with unemployment benefits (the Executive Institute for Employees Insurances). In all articles, four indicators of blame attribution were used and presented in a similar way. All in all, we ensured that the wording of the stimuli only differed on the independent variables manipulated in this study. Manipulation Checks A pilot test among a convenience sample (N = 117) indicated that the manipulations were successful and that the stimuli affected perceptions of blame and emotional styles in the predicted direction. The manipulation check items used in the main study asked participants to remember the causes of the worsening Labour market situation mentioned in the newspaper article. The manipulation of causal responsibility attribution was successful for both blame attributed to the EU (F(2, 707) = 75.40, p <.001) and the national government (F(2, 709) = 49.44, p <.001). The manipulation of emotional style succeeded as well (F(2, 713) = 15.01, p <.001). Measures Unless explicitly stated otherwise, all items were measured on a 7-point scale (1 = completely disagree, 7 = completely agree). All items are included in their exact wording in Appendix B.

11 880 Communication Research 44(6) Citizens perceptions of blame. The key dependent variable was measured in two different ways. First, similar to Iyengar s (1991) measure of responsibility attribution, perceptions of blame were measured with an open-ended question. Just like Iyengar asked participants to indicate what the most important causes of the issue were, we posed the following question: Can you describe whom or what you feel is most responsible for causing this situation? The responses to this question were recoded into four categories (1 = blame assigned to Dutch government, 2 = blame assigned to the EU, 3 = blame assigned to others, 4 = blame assigned to both the EU and national government). Intercoder reliability was assessed for a sample of 141 (19.6%) randomly selected open-ended questions coded by two independent coders who were trained by a detailed coding procedure that explained each of the four categories in detail. Krippendorff s alpha was.86. Second, participants were presented with 19 different candidates for responsibility (Gomez & Wilson, 2008). These different actors, organizations, political elites, and other out-groups appeared in a random order. Participants were asked to what extent they believed each was responsible for the worsening Labour market situation on a 7-point scale (1 = completely responsible, 7 = not at all responsible). Populist attitudes. Following definitions by Mudde (2004) and Jagers and Walgrave (2007), populism consists of two core components: the (failed) representation of the ordinary people and the moral antagonism between the good people and the evil elites. Confirmatory Factor Analysis (CFA) was used to test whether the data fitted this twofactor measurement model (items derived from Akkerman, Mudde, & Zaslove, 2014; Hawkins, Riding, & Mudde, 2012). The model fitted the data well (χ 2 (5) = 10.85, p =.054; root mean square error of approximation [RMSEA] = 0.023, 90% confidence intervals [CI] = [0.00, 0.04]; comparative fit index [CFI] = 0.99). The correlation between the factors was.58. To further test for discriminant validity, we constrained the correlation between the two factors to 1.00, which resulted in a significant decline in model fit ( χ 2 = 30.06, p <.001). This indicates that although both factors were related, they correlated too weakly to be merged into a one-dimensional populist attitudes scale. Two 7-point scales were constructed: Representation (M = 4.81, SD = 1.19, Cronbach s α =.76) and Antagonism (M = 4.04, SD = 1.19, Cronbach s α =.58). Higher scores on both scales indicate stronger populist attitudes on the corresponding dimension. The items for the two dimensions of populist attitudes are listed in Appendix B. The items measuring the representation scale tap into participants perception that politicians should respond to the lifeworld of ordinary citizens by listening to their concerns, by understanding the problems they are facing, and by following their will. The essence of this scale thus describes the ways in which politicians, as representatives of the people, should be responsive to the problems experienced by the people (e.g., Albertazzi & McDonnell, 2008). The items of the antagonism scale tap into a perceived divide between the good ordinary people and the bad politicians. This scale thus vertically constructs political elites as the populist other (e.g., Hawkins, 2010; Mudde, 2004).

12 Hameleers et al. 881 Moderators: Identity Attachment to the Netherlands and Europe We used two three-item scales to measure social identity attachment. The items were based on measures used by Lubbers (2008) and Boomgaarden, Schuck, Elenbaas, and de Vreese (2011). We constructed a scale for Dutch identity attachment (M = 5.43, SD = 1.26, Cronbach s α =.92) and a scale for European identity attachment (M = 3.84, SD = 1.49, Cronbach s α =.86). Higher scores on the 7-point scales indicate a stronger identity attachment. Control Variables Our design allowed us to assure random assignment over experimental conditions, and we ensured that assignment to specific experimental conditions did not result in differential attrition (Mutz & Pemantle, 2011). For illustrative purposes, we performed a between-conditions randomization check on the control variables age, gender, level of education, news exposure, political efficacy, political distrust/cynicism, voting behavior, and attitudes toward the Labour market situation. We found no significant differences between experimental conditions regarding these variables. 3 Analysis Table B1 in Appendix B shows descriptive statistics of the main dependent variables for all experimental conditions. To test the hypotheses, logistic regression analysis and multivariate analyses of covariance (MANCOVAs) 4 were conducted to assess how different levels of emotionalized blame attributions affected citizens blame perceptions and populist attitudes. In addition, multiple linear regression was used to assess how identity attachment moderated the effects of emotionalized style on out-group perceptions. Results Direct Effects of Populist Blame Attributions The direct effects of populist attributions of blame are graphically depicted in Figure 1. First, we conducted a logistic regression analysis to assess how populist blame attributions affected open-ended blame perceptions. Blame attribution significantly affected open-ended blame perceptions toward both the Dutch government (b = 1.70, SE = 0.18, p <.001) and the EU (b = 3.05, SE = 0.28, p <.001) in the hypothesized direction. The odds ratio for the governmental blame attribution coefficient was 5.48 (95% CI = [3.86, 7.79]), which indicates that citizens in the governmental blame conditions were 5.48 times more likely to blame the government than citizens in the no-blame or EU-blame conditions. The odds ratio for the EU blame attribution coefficient was (95% CI = [12.10, 36.90]), which indicates that citizens in the EU blame conditions were times more likely to blame the EU than citizens in the no-blame attribution or

13 882 Communication Research 44(6) EU Blame Perceptions Government Blame Perceptions Populist attitude: Representation 3 No Blame EU Blame Government Blame Populist attitude: Antagonism Figure 1. Direct effects of populist blame attributions on dependent variables. Table 2. Scores on Dependent Variables for Different Blame Attribution Conditions. Experimental condition No-blame attribution EU blame attribution Government blame attribution Direct effects: F, Df (2, 696) Partial η 2 EU blame perceptions 4.39 a (1.28) 4.78 b (1.23) 4.33 c (1.31) 8.53** 0.02 Government blame 4.69 a (1.31) 4.86 a (1.23) 4.78 a (1.33) 0.18 perceptions Pop 1: Representation 4.63 a (1.15) 4.90 b (1.07) 4.97 b (1.12) 5.32** 0.02 Pop 2: Antagonism 3.96 a (1.04) 4.15 a (1.10) 4.07 a (1.18) 0.62 Note. Standard deviations are reported in parentheses to the right of means. Means with differing subscripts within the rows differ significantly at the p <.05 level based on independent samples t-test. *p <.05. **p <.01. governmental blame attribution conditions. The answers to the closed-ended questions only provide support for a direct effect of populist blame attribution on blame perceptions toward the EU (F(2, 696) = 8.53, p <.001, partial η 2 = 0.02). The effect again pointed in the hypothesized direction (see Table 2 for mean score differences between experimental conditions). Of the two populist attitudes, only the representation dimension was directly affected by populist attributions of blame. Specifically, citizens exposed to attributions of responsibility to the EU or the national government were more inclined to believe politicians failed to represent the people s will than citizens exposed to no-responsibility attributions (see Table 2 for mean differences). Overall, the results offer partial support for Hypothesis 1. As expected, populist blame attributions were taken over by citizens but only affected both open-ended and

14 Hameleers et al. 883 closed-ended measures of blame when the EU was attributed responsibility. Moreover, only the populist representation attitude was directly affected by populist blame attributions. The Perceptual Screen of Identity Attachment The results indicate that identity attachment functions as a perceptual screen when citizens respond to populist blame attributions. We found a significant two-way interaction effect of blame attribution and Dutch identity attachment on citizens governmental blame perceptions: F(2, 670) = 5.38, p =.005, partial η 2 = As shown in Table 3, citizens experiencing a weaker attachment to national identity accepted blame frames that attributed responsibility to the Dutch government. In contrast, citizens experiencing a stronger attachment to national identity were not persuaded by populist blame frames (see Table 3 for mean differences between experimental conditions). These findings offer support for H2a. As support for the national government or partisanship provides an important alternative explanation of the perceptual screen mechanism moderating the effects of blame attribution, we also estimated the role of support for the government as robustness check. The results of this analysis indicated that participants perceptual screen was only based on national identity and not on support for the government, 5 which further supports H2a. The two-way interaction effect of responsibility attribution and European identity attachment on EU blame perceptions was nonsignificant (F(2, 669) = 0.059, p = n.s.), which is not supportive of H2b. The interaction effect of responsibility attributions and national identity attachment on the populist antagonism attitude approached statistical significance (F(5, 715) = 2.16, p =.057, partial η 2 = 0.02) and the interaction effect on the representation attitude was highly significant (F(5, 715) = 7.31, p <.001, partial η 2 = 0.05). In support of H2a, the populist attitudes of citizens with a weaker national identity attachment were stronger in the blame conditions than in the no-blame conditions (see Table 4 for mean differences). As expected, differences between blame attribution conditions were not significant for citizens who identify strongly with the nation state. We found significant interaction effects of blame attributions and European identity attachment on the populist representation attitude (F(5, 715) = 16.28, p <.001, partial η 2 = 0.10) and the populist antagonism attitude (F(5, 715) = 8.03, p <.001, partial η 2 = 0.05). In support of H2b, populist attributions of blame resulted in stronger populist attitudes on both dimensions among the people with a weaker attachment to Europe. Overall, the results provide support for Hypothesis 2. With the exception of blame perceptions toward the EU, identity attachment functioned as a perceptual screen moderating the effects of populist blame attributions on perceptions toward the political establishment.

15 Table 3. Effects of Emotionalized Blame Attributions on Blame Perceptions for Different Levels of Identity Attachment. Experimental conditions Dependent variable Identity attachment level No-blame attribution EU blame attribution Government blame attribution Anger Fear Total Anger Fear Total Anger Fear Total Three-way interaction effect F, Df (2, 681) Partial η 2 Perceptions EU blame Perceptions Government blame Weaker EUidentity Stronger EUidentity 4.66 a,x 4.15 b,x 4.44 c,x 4.90 a,x 5.02 d,x 4.95 e,x 4.80 a,x 4.53 a,x 4.67 c,x 3.07** 0.03 (1.35) (1.34) (1.37) (1.07) (1.46) (1.25) (1.42) (1.43) (1.43) 4.36 a,x 4.50 b,x 4.34 c,x 4.50 a,x 4.79 b,x 4.66 d,x 4.01 a,y 4.12 a,x 4.07 c,y (1.10) (1.24) (1.21) (1.20) (1.20) (1.20) (1.26) (1.02) (1.14) Total 4.46 a 4.33 b 4.36 c 4.69 a 4.89 d,x 4.79 e 4.36 a 4.27 a 4.31 c (1.22) (1.30) (1.28) (1.16) (1.28) (1.22) (1.37) (1.24) (1.31) Weaker national identity Stronger national identity 4.77 a,x 4.02 b,x 4.28 c,x 5.21 a,x 5.56 d,x 5.37 e,x 5.35 a,x 3.32 a,x 4.92 c,z 2.14* 0.02 (1.48) (1.68) (1.55) (1.50) (1.05) (1.29) (1.39) (1.50) (1.40) 4.99 a,x 4.79 b,x 4.74 c,y 4.76 a,x 4.86 b,x 4.81 c,y 4.74 a,x 4.82 a,y 4.78 c,x (1.11) (1.33) (1.28) (1.22) (1.21) (1.21) (1.39) (1.22) (1.30) Total 4.93 a 4.72 b 4.67 c 4.79 a 4.92 b 4.85 c 4.81 a 4.73 a 4.77 c (1.15) (1.34) (1.30) (1.26) (1.21) (1.21) (1.38) (1.27) (1.33) Note. Standard deviations are reported in parentheses below means. Means with differing first subscripts (a through e) within the rows and within emotionalized style differ significantly at the p <.05 level based on independent samples t-test comparing no-blame attribution with EU or Dutch government blame attribution conditions. Means with differing second subscripts (x through z) within columns indicate significant differences between levels of identity attachment within experimental conditions. Identity attachment scales on the national and European level were recoded into weaker and stronger in the same way. The threshold value for higher attachment was set at Analyses with alternative recoding yielded similar results. *p <.05. **p <

16 Table 4. Effects of Emotionalized Blame Attributions on Populist Attitudes for Different Levels of Identity Attachment. Experimental conditions Dependent variable Identity attachment level No-blame attribution EU blame attribution Government blame attribution Anger Fear Total Anger Fear Total Anger Fear Total Three-way interaction effect F, Df (2, 681) Partial η 2 Populist attitude: representation Populist attitude: antagonism Weaker national identity Stronger national identity 5.06 a,x (1.19) 4.44 a,x (1.11) Total 4.73 a (1.19) Weaker national identity Stronger national identity 3.52 a,x (1.04) 4.23 a,y (0.99) Total 4.16 a (1.02) 4.74 b,x (1.13) 4.30 b.x (1.23) 4.51 b (1.21) 3.89 b,x (1.49) 3.89 b,x (1.00) 3.87 b (1.05) 4.92 c,x (1.09) 4.33 c,y (1.10) 4.63 c (1.15) 3.88 c,x (1.23) 3.97 c,x (1.02) 3.94 c (1.04) 5.42 a,x (1.01) 4.66 a,y (1.22) 5.02 a (1.18) 4.64 d,x (1.32) 4.13 a,x (1.06) 4.19 a (1.10) 5.33 d,x (1.27) 4.52 b,y (1.13) 4.80 b (1.23) 4.77 b,x (0.92) 4.05 b,y (1.12) 4.13 b (1.11) 5.38 e,x (1.13) 4.58 c,y (1.17) 4.91 d (1.21) 4.70 e,x (1.13) 4.09 c,y (1.09) 4.16 c (1.11) 5.36 a,x (1.19) 4.53 a,y (1.18) 4.94 a (1.25) 4.89 f,x (1.32) 4.07 a,y (1.16) 4.19 a (1.21) 5.31 d,x (1.11) 4.70 d,x (1.08) 5.00 e (1.13) 2.50 b,x (1.00) 4.01 b,y (1.13) 3.93 b (1.13) 5.34 e,x (1.15) 4.62 c,y (1.13) 4.97 f (1.19) 4.39 c,x (1.59) 4.04 c,x (1.14) 4.06 c (1.18) 7.25** ** 0.03 Note. Standard deviations are reported in parentheses below means. Means with differing first subscripts (a through f) within the rows and within emotionalized style differ significantly at the p <.05 level based on independent samples t-test comparing no-blame attribution with EU or Dutch government blame attribution conditions. Means with differing second subscripts (x and y) within columns indicate significant differences between levels of identity attachment within experimental conditions. Identity attachment scales on the national and European level were recoded into weaker and stronger in the same way. The threshold value for higher attachment was set at Analyses with alternative recoding yielded similar results. *p <.05. **p <

17 886 Communication Research 44(6) Effects of Emotional Blame Attributions on Blame Perceptions and Populist Attitudes We found significant two-way interaction effects of emotionalized style and blame attributions on blame perceptions toward the EU (F(2, 681) = 3.65, p =.026, partial η 2 = 0.01) and the Dutch government (F(2, 681) = 3.17, p =.043, partial η 2 = 0.01). In line with H3a, citizens in the fear conditions were more likely to accept blame attributions to the EU than citizens in the anger conditions (see Table 3 for mean differences between experimental conditions). When blame was attributed to the government, fear also resulted in more acceptance of the blame frame than anger, but the mean differences were nonsignificant. The significant two-way interaction effect of blame attributions and emotional style on the populist representation attitude (F(2, 681) = 2.74, p =.012, partial η 2 = 0.02) indicates that citizens in the fear conditions were significantly more likely to believe that politicians should, but failed to, represent the ordinary people s will as a result of exposure to governmental blame attributions than citizens in the anger conditions (see Table 4 for mean differences). Emotionalized attributions of blame to the EU, however, did not affect any of two populist attitudes. The results are predominantly supportive of H3a. As predicted, fear resulted in a stronger tendency to accept populist blame attributions than anger. However, this effect was only significant when blame was attributed to the EU. Populist attitudes, however, were only significantly affected by fear-emphasizing blame attributions to the government. Identity Attachment as a Moderator of Emotional Blame Attribution Effects In a next step, regression analyses showed how the effects of the emotional style of the message were contingent upon different levels of identity attachment. As shown in Table 5, the effects of the emotional style on governmental blame perceptions were moderated by national identity attachment. For lower levels of national identity attachment, anger resulted in stronger governmental blame perceptions than fear. In contrast, at higher levels of national identity attachment, fear resulted in stronger blame perceptions than anger (see Figure 2 for the regression lines of the interaction effects). Identity attachment to Europe did not change the relationship between emotional style and EU blame perceptions, as demonstrated by the nonintersecting lines in Figure 2. The regression analysis points to a significant interaction effect of national identity attachment and emotional style on the populist representation attitude. For lower levels of identity attachment, anger resulted in stronger perceptions of failed representation than fear. For higher levels of identity attachment, the effect was the other way around. We also found a significant interaction effect of European identity attachment and emotionalized style on the populist antagonism attitude. As expected, fear resulted in stronger good people versus corrupt elites perceptions than anger for lower levels of European identity attachment (see Figure 2).

18 Hameleers et al. 887 Table 5. Regression Model Predicting the Effects of Emotionalized Style at Different Levels of Identity Attachment. Dependsent measures Variable Model I Model II B (SE B) B (SE B) Governmental blame perceptions EU blame perceptions Populist attitude: representation Populist attitude: antagonism (Constant) 4.84 (0.13)** 4.83 (0.13)** Emotionalized style 0.10 (0.18) 0.13 (0.18) National identity 0.13 (0.07) 0.14 (0.09) Emotionalized style 0.41 (0.15)** National identity Adjusted R F for change in R ** (Constant) 4.66 (0.12)** 4.66 (0.12)** Emotionalized style 0.22 (0.17) 0.22 (0.17) European identity 0.13 (0.05)* 0.15 (0.08) Emotionalized style 0.05 (0.11) European identity Adjusted R F for change in R 2.25 (Constant) 4.93 (0.12)*** 4.92 (0.11)*** Emotionalized style 0.09 (0.17) 0.04 (0.16) National identity 0.12 (0.07) 0.26 (0.08)** Emotionalized style 0.46 (0.14)** National identity Adjusted R F for change in R ** (Constant) 4.16 (0.11)*** 4.17 (0.11)*** Emotionalized style 0.01 (0.15) 0.02 (0.15) European identity 0.13 (0.05)* 0.02 (0.08) Emotionalized style 0.19 (0.10)* European identity Adjusted R F for change in R * Note. Effects are analyzed within blame attribution conditions. Government blame and representation attitudes are analyzed within governmental blame attribution conditions (n = 224). EU blame and populist antagonism attitudes are analyzed within EU blame attribution conditions (n = 223). National identity and European identity were centered at their means. Standard errors are reported between parentheses. *p <.05. **p <.01. ***p <.001. In sum, the results offer limited support for Hypothesis 3b. In support of H3b, fear resulted in a stronger perceived antagonism between the good people and corrupt elites than anger for citizens with a weaker attachment to Europe. In contrast to H3b, fear did not result in stronger populist attitudes and governmental blame perceptions for lower levels of national identity attachment. But how is the acceptance of blame frames contingent upon the emotional style for different levels of identity attachment?